Create functions for composing nodeadm and bottlerocket user data

[flostadler]

Hello!

• Vote on this issue by adding a 👍 reaction
• If you want to implement this feature, comment to let us know (we'll work with you on design, scheduling, etc.)

Issue details

AL2023 and Bottlerocket use structured configuration for the node and kubelet. This configuration is passed to the instance via the user data document (i.e. a string).

Status Quo
AL2023 configuration is a MIME multipart/mixed document and can be interleaved with arbitrary data and bash scripts. Because of this we cannot reasonably expose it as typed data within the node groups. Instead it's an array of multi part sections:

{
    content: pulumi.Input<string>;
    contentType: pulumi.Input<string>;
}

Bottlerocket configuration has to be in TOML format. Because of that we can model it as a object. But it's lacking type information of the expected structure.

Proposal
We can add functions to compose the expected NodeConfig yaml for nodeadm (i.e. AL2023) and bottlerocket's TOML config.

One challenge here will be forward and backwards compatibility. We'll have to namespace the configuration like we do for kubernetes resources
E.g. eks.nodeadm. v1alpha1.getNodeConfig({...}).

Bleeding edge users can leverage the untyped properties directly as an escape hatch in case the provider does not yet provide updated types for new versions.

Another challenge that needs to be solved is generating the types. Both nodeadm and bottlerocket provide schemas (openapi schema or k8s CRD), but their types are lacking the information we need. They're modeled as object as well (e.g. see).

Affected area/feature